Verbal Index of Architectural Complexity (VIAC)

In the MAS perspective, intelligence is treated as a meta-mode of a meta-algorithmic system, where problem-solving relies on multiple levels of models and meta-operations. A natural structural descriptor of such a regime is architectural depth: the number and quality of actually functioning modeling/control layers.

Text is treated as a projection of that architecture into a language layer: words act as markers of fragments of a conceptual field of abstractions, which can be approximated as a lexico-semantic graph. Within this view, VIAC aggregates an estimate of the “average abstraction height” expressed in a text or corpus.

The key claim is intentionally one-sided: sustained expression of high-level abstraction patterns in a sufficiently rich corpus constrains the minimum architectural depth required for such production to be stable. VIAC is therefore positioned as an architectural floor, not a complete measure of intelligence.

Many established metrics quantify readability or statistical predictability. They can be useful, but they do not target the ontological level of abstractions used by the generating system. VIAC is designed to operate on a different axis: not “hard to read”, but “how high in abstraction the content reliably lives”, in a way compatible with the MAS picture.

VIAC is not an IQ surrogate. It does not depend on psychometric tasks and is not anchored to a human-specific benchmark; it is intended as a substrate-agnostic tool for comparing heterogeneous cognitive architectures when a verbal trace is available.

The project is currently developed as a theoretical scaffold and a research protocol. Implementation details, calibration choices, and evaluation procedures are disclosed selectively after methodological stabilization.